sn architecture

[Aurelian Melinte <ame01@gmx.net>]

hello,

over the past two months I worked on a Java parser for SN, to replace the 
original one. More on that below. During this period I had some suggestions 
from more knowledgeable peoples that I think would be beneficial to SN.

a) It would be good to add to the API libraries functions that will allow 
to query the library from within  a parser. Example: if I have a function call
     somemethod( par1, par2 );
and par1 is of type class P from which classes C1 and C2 are inheriting, 
then par1 can be of type P or C1 or C2. But when I parse Java file by file, 
I have no way to know it unless I query the database. This will improve the 
cross-ref functionality to better display polymorphism.

b) It would be good to turn the code of those libraries into loadable 
modules (.dll, .so, etc). Thus, I can supply my own without having to 
recompile all the parsers. The idea behind is that I will be able to add to 
the current structure of the database transparently, to store more 
information, to support different kinds of queries, etc.  Or, I can use my 
own kind of database just by changing this module.

c) In this case, the tcl/tk code probably would need to me modified to 
query through this module too, and, eventually, group the query 
functionality in one place. It would be easier to modify it - I noticed 
that although I can insert xrefs to classes or to local variables, those 
are not displayed. I have no intimate knowledge with tcl, sorry.



About the replacement Java parser: you can find it at 
http://209.47.210.117:8080/sn.html for now. It detects interface/class 
inheritance, interface/class structure, methods and class variables 
properly. It detects some cross-references too but very limited (I will 
resume work on it in a couple of weeks probably). It is much slower than 
the original :(. It requires JDK/JRE to work.

Regards,
a.